sub/sd_ass: subtitle timing rounding error

[kako94321]

Use nearbyint() when the value is extremely close to an integer and floor() otherwise to prevent forward rounding caused by floating-point precision errors. This avoids early subtitle presentation on boundary frames.

Add test for ASS subtitle timing to ensure correct rendering around boundary timestamps.

Fixes: #17453

Read this before you submit this pull request:
https://github.com/mpv-player/mpv/blob/master/DOCS/contribute.md

Reading this link and following the rules will get your pull request reviewed
and merged faster. Nobody wants lazy pull requests.

[kasper93]

@arch1t3cht: Could you do the same in aegi? floor(pts * 1000 + 1e-6); or something? I think it would be prudent to align on the solution. Adding epsilon fixes, the issue mentioned here #17453 (comment). (i.e. floor(518.81 * 1000 + 1e-6) = 518810)

[kasper93]

Please squash the commits. I also think that a test for single math operation is a bit overkill.

[kako94321]

Of course. I'm removing the test file before squashing then

[kasper93]

LGTM, thanks!

[arch1t3cht]

Could you do the same in aegi?

Yeah, I will. A hardcoded + 1e-6 does seem a bit hacky to me but there's no real better fully satisfying solution (except reworking the entire pipeline to never use floats lol) and I'm not sure how much it's worth bikeshedding this further.

[kasper93]

In fact there are 3 other places where timestamp is converted. Could you update them too?

[cubicibo]

Why not just use nextafter(double, double) prior to flooring?

[kasper93]

Why not just use nextafter(double, double) prior to flooring?

Having bigger epsilon is safer. If the pts accumulated more error across it's lifetime, it's not certain nextafter will be enough for this.

[kako94321]

In fact there are 3 other places where timestamp is converted. Could you update them too?

Sure. Are you also referring to llrint(pkt->duration * 1000)?

[kasper93]

In fact there are 3 other places where timestamp is converted. Could you update them too?

Sure. Are you also referring to llrint(pkt->duration * 1000)?

Yes, and 2 more. Search for llrint in that file.